شیرین امینی

Today, with the popularity of social media platforms, people use sarcasm when expressing their opinions on various topics, products, and services, including political events, sports games, and daily activities and posts. Sarcasm is a form of irony and occurs when there is a discrepancy between the literal meaning and the intended meaning of a statement. Sarcasm is ubiquitous in social media context and can disrupt systems that use this data to analyze sentiment and sentiment. The purpose of this research is to detect sarcasm with Burt's algorithm and compare it with other models. Detecting sarcasm, the proposed method of the thesis consists of two stages: in the first stage, pre-processing of the texts is done. In the second stage, sarcasm is detected with machine learning and deep learning approaches. In classical machine learning models, term frequency-inverse frequency (TF_IDF) method is used for text representation. And in the Brett model, the Brett embedding is used, and in other models, the embedding suitable for each model is used. The proposed work is evaluated on three datasets that include news headlines and tweets. The new headline dataset includes news headlines, the Isarcasm dataset, and the Twitter dataset includes tweets. The results show that Burt's algorithm with ۹۳% and ۸۶% accuracies respectively on the set of new headlines and Isarcasm and Roberta's model with ۹۳% accuracy on the Twitter set have obtained the best accuracy compared to other models.